计算机工程与应用 ›› 2007, Vol. 43 ›› Issue (8): 142-144.

• 数据库与信息处理 • 上一篇    下一篇

基于多个混合核函数的SVM决策树算法设计

薛欣 贺国平   

  1. 山东科技大学 山东科技大学
  • 收稿日期:2006-07-14 修回日期:1900-01-01 出版日期:2007-03-11 发布日期:2007-03-11
  • 通讯作者: 薛欣

Designing the Algorithm of SVM Decision Tree
Based on Many a Mixture of Kernels

XUE Xin 1,2 HE Guoping1

  

  1. .College of Info. Sci. & Eng.,SUST, Qingdao Shandong 266510,China;
    2.Dept. of Math., Taishan University, Taian Shandong 271021,Chin
  • Received:2006-07-14 Revised:1900-01-01 Online:2007-03-11 Published:2007-03-11

摘要: 不同的核函数具有不同的特性,SVM决策树中每个子SVM面对的分类对象不同,选取的核函数及其参数也应该不同。通过调节混合核函数的参数 形成不同的核函数,给出了一个用多个混合核函数训练SVM决策树的多类分类算法。仿真试验表明,该算法与只用一个核函数训练SVM决策树的算法相比,具有较高的分类精度。

关键词: 核函数, SVM决策树, 混合核函数

Abstract: Different kernel function has different characters. Every sub-SVM should have different kernel in SVM decision tree. An algorithm of SVM decision tree based on many a mixture of kernels is given. Experimental results show that the algorithm of SVM decision tree based on different kernels has higher force of classification than that with only one kernel.

Key words: kernel function, SVM decision tree, mixture of kernels